AI in healthcare: Inflationary today, deflationary tomorrow | Viewpoint

Healthcare leaders are asking a deceptively simple question: Will AI increase costs or reduce them? The honest answer is both. However, the distinction comes down to timing, incentives, and execution.

In the near term, AI is proving to be inflationary. Organizations are investing heavily in infrastructure, data platforms, and new applications. At the same time, many of the earliest scaled use cases—ambient documentation, revenue cycle optimization, and payment integrity—are designed to improve financial performance within existing models, not fundamentally reduce the total cost of care. That distinction matters.

Take AI-assisted documentation as an example. A study published in JAMA Network Open found that ambient AI scribes were associated with a 5.8% increase in weekly relative value units (RVUs). In plain terms, that means doctors using these tools generated roughly 6% more billable clinical work each week—evidence of inflationary AI.

Across the industry, there are similar signals: higher risk recapture, more complete risk coding, and more thorough documentation of clinical complexity. In many cases, this reflects a more accurate representation of patient conditions, not inappropriate behavior. But the financial impact is clear. But reimbursement systems that tie payment to documented risk and complexity can still create inflationary effects. More documented complexity often translates into higher reimbursement and, in aggregate, higher spending.

At the same time, AI is lowering the barrier to executing work that was previously constrained by human capacity. When it becomes easier to scale outreach, documentation, or coding, activity increases. That can drive short-term cost expansion before efficiency gains begin to materialize.

This is not a failure of AI. It reflects where we are applying it. Right now, most implementations are focused on optimizing the existing system. They are making current workflows faster and more scalable, but they are not yet fundamentally changing how care is delivered. That is where the long-term opportunity lies.

The shift from optimization to transformation

Over time, AI has the potential to be deflationary, but only if it is applied beyond administrative and financial workflows. The real opportunity is in changing when and how care happens.

Healthcare has historically operated in a reactive model. Care management, for example, has largely been “damage control,” only intervening after a hospitalization or once a chronic condition has already escalated. The constraint has always been capacity. Teams simply do not have the bandwidth to continuously monitor and engage entire populations. AI begins to change that equation.

By analyzing clinical, behavioral, and utilization patterns across large populations, AI can surface early signals of risk and trigger timely intervention. Instead of waiting for a patient to become high-cost or high-acuity, organizations can act earlier—often at a lower cost and with better outcomes. This is where AI starts to bend the cost curve.

We are already seeing early examples. Tasks that once required large teams, such as outbound patient engagement or follow-up coordination, can now be executed more efficiently with AI-supported workflows. What used to take hours across multiple staff members can now be completed in a fraction of the time, often with greater consistency. That’s because AI unlocks capacity and scale.

The same applies to care transitions. Patients moving between care settings represent one of the highest-risk and highest-cost moments in the system. Missed follow-ups, medication confusion, and gaps in communication drive avoidable utilization. AI can help identify these transition points in real time and coordinate interventions that improve continuity and reduce downstream costs.

Medication adherence is another area with significant potential. When patients better understand their treatment plans and receive timely reminders or support, adherence improves. That, in turn, supports disease control, reduces complications, and combats avoidable utilization.

These are not theoretical benefits. They are practical applications that begin to shift healthcare from reactive to proactive.

The workflow problem behind the cost problem

There is a broader issue that sits beneath this discussion: healthcare does not have a technology problem. It has a workflow problem.

Many AI tools generate insights, but those insights often live outside the environments where decisions are made. If an insight is not delivered at the point of care, embedded within the clinician or care team’s workflow, it is unlikely to drive action. This is one of the biggest barriers to realizing AI’s deflationary potential.

There are already early examples of how embedding AI directly into operational workflows can compress decision-making cycles and improve organizational responsiveness. In one case, conversational analytics reduced iteration cycles between analysts and executives by as much as 50%, often shortening project turnaround times by up to a full week. Instead of waiting days for follow-up analyses or additional reporting, leaders were able to ask questions conversationally and receive data-backed insights in near real time.

Embedding AI into everyday decision-making is not trivial. It requires integration with clinical systems, alignment with operational processes, and trust from frontline users. It is also a significant change management effort. Organizations are not just deploying new tools; they are rethinking how work gets done.

That distinction is important because the long-term value of AI will not come solely from automating existing tasks. It will come from reducing friction across the decision-making process itself – allowing organizations to move faster, intervene earlier, and operate with greater precision at scale.

That is why progress can feel slower than expected. It is not just about implementing AI. It is about operationalizing it. Until insights are consistently translated into action, the impact on cost and outcomes will remain limited.

Policy tailwinds and the role of incentives

The shift toward deflationary impact is also tied to incentives. Emerging models from The Centers for Medicare & Medicaid Services, including the ACCESS model, are increasingly aligning reimbursement with outcomes rather than volume. That changes the equation for how AI is applied.

When organizations are accountable for total cost of care and patient outcomes, the value of early intervention, proactive engagement, and coordinated care becomes much clearer. AI is not just a tool for optimization, as it is becoming a lever for performance.

This alignment is critical. Without it, there is a risk that AI continues to drive efficiency within siloed functions like improving revenue cycle performance or documentation accuracy, without materially changing overall spending. With it, the focus shifts toward interventions that reduce avoidable utilization and improve long-term outcomes.

Two forces that will drive long-term deflation

Looking ahead, two dynamics are likely to shape AI’s long-term impact on healthcare costs.

The first is administrative deflation. As AI takes on routine, repetitive tasks, organizations can streamline administrative workflows and free up resources for higher-value work. This is not about eliminating roles, it is about changing how teams spend their time. Less effort on documentation and coordination, more focus on clinical decision-making and patient care.

The second is patient activation. As individuals gain access to better tools, clearer information, and more timely support, they can play a more active role in managing their health. This is particularly important for chronic conditions, where more informed, engaged patients have better outcomes and lower long-term costs.

This dynamic is especially important in rural and underserved communities, where capacity and access constraints are most acute. AI can help extend the reach of existing teams, enabling more consistent engagement with broader swaths of the population without requiring proportional increases in staffing.

The bottom line

AI will not reduce healthcare costs overnight. In fact, in the near term, it may do the opposite. But focusing only on short-term financial impact misses the bigger picture. The real question is not whether AI is inflationary or deflationary. It is whether we are applying it in ways that fundamentally change how care is delivered.

If AI remains concentrated in administrative and financial workflows, its impact will be incremental. If it is embedded into clinical and operational workflows—like supporting earlier intervention, better coordination, and more engaged patients—it has the potential to reshape both outcomes and cost structures over time.

The technology is advancing quickly. The challenge now is execution. And in healthcare, execution is what ultimately determines whether innovation translates into impact.

Michael Meucci is president and CEO of Arcadia